inst/doc/stop-identification.R

In stdbscan: Spatio-Temporal DBSCAN Clustering

## ----include = FALSE----------------------------------------------------------
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

## ----setup, message=FALSE-----------------------------------------------------
library(stdbscan)
library(readr)
library(ggplot2)
library(plotly)

## -----------------------------------------------------------------------------
head(geolife_traj)

## ----fig.wigeolife_trajh=7, fig.height=7--------------------------------------
ggplot() +
  geom_path(data = geolife_traj, aes(x, y)) +
  labs(x = "", y = "",
    title = "GPS track analyzed in this vignette",
    caption = "Data: GeoLife GPS Trajectories (Microsoft, 2012). Author: Antoine Le Doeuff, 2026",
  ) +
  coord_equal() +
  theme_minimal() +
  theme(plot.title = element_text(size = 16, face = "bold"))

## -----------------------------------------------------------------------------
geolife_traj$date_time <- as.POSIXct(
  paste(geolife_traj$date, geolife_traj$time),
  format = "%Y-%m-%d %H:%M:%S",
  tz = "GMT"
)

# Sort data by time if needed
geolife_traj <- geolife_traj[order(geolife_traj$date_time), ]

# Convert to cumulative time
geolife_traj$t <- as.numeric(
  geolife_traj$date_time - min(geolife_traj$date_time)
)

# Convert to matrix
data <- cbind(geolife_traj$x, geolife_traj$y, geolife_traj$t)

## ----fig.height=8-------------------------------------------------------------
(res <- st_dbscan(
  data = data,
  eps_spatial = 3, # meters
  eps_temporal = 30, # seconds
  min_pts = 3,
  # extra arguments
  splitRule = "STD",
  search = "kdtree",
  approx = 1
))

## ----fig.wigeolife_trajh=7, fig.height=7--------------------------------------
# Put the cluster in the input data
geolife_traj$clust <- as.factor(res$cluster)

# Extract stops and movements
geolife_traj_mvt <- geolife_traj[geolife_traj$clust == "0", ]
geolife_traj_stop <- geolife_traj[geolife_traj$clust != "0", ]

# Plot
ggplot() +
  geom_path(data = geolife_traj_mvt, aes(x, y)) +
  geom_point(data = geolife_traj_stop, aes(x, y, color = clust), size = 4) +
  labs(x = "", y = "", color = "stop ID",
    title = "ST-DBSCAN stop identification",
    subtitle = "eps_spatial = 3 m, eps_temporal = 30 s and min_pts = 3",
    caption = "Data: GeoLife GPS Trajectories (Microsoft, 2012). Author: Antoine Le Doeuff, 2026",
  ) +
  scale_color_manual(values = MetBrewer::met.brewer("Isfahan2", 5)) +
  coord_equal() +
  theme_minimal() +
  theme(
    legend.position = "bottom",
    plot.title = element_text(size = 16, face = "bold"),
  )

## ----warning=FALSE, message=FALSE---------------------------------------------
# Zoom on stop 4
geolife_traj_f <- geolife_traj[
  geolife_traj$x > 441060 & geolife_traj$x < 441100,
]
geolife_traj_f <- geolife_traj_f[
  geolife_traj_f$y > 4428780 & geolife_traj_f$y < 4428820,
]

# Extract stop
geolife_traj_f_stop <- geolife_traj_f[geolife_traj_f$clust != "0", ]

# Plotly figure
fig <- plot_ly(
  data = geolife_traj_f,
  x = ~x,
  y = ~y,
  z = ~t,
  type = "scatter3d", mode = "lines+markers",
  line = list(wigeolife_trajh = 4, color = "grey"),
  marker = list(size = 3, color = "grey")
)
fig |>
  add_markers(
    x = ~geolife_traj_f_stop$x,
    y = ~geolife_traj_f_stop$y,
    z = ~geolife_traj_f_stop$t,
    marker = list(size = 4, color = 'red'),
    name = 'Stop'
  ) |>
  layout(
    scene = list(
      xaxis = list(title = "x"),
      yaxis = list(title = "y"),
      zaxis = list(title = "t")
  )
)